MenuStart a projectStart

AI & data

AI-narrated audio tours with ElevenLabs

This automation turns the notes your team writes for each stop on a tour into a spoken script and a narrated recording. After a person has approved the script and listened to the audio, it is published with a transcript on a page visitors reach by scanning a QR code.

An example of how the studio would build it. No live services are connected.

From your own notes to an audio tour visitors can play.

  1. Step 01

    Notes written

    A member of the team writes or updates the notes for one stop, such as a room, an exhibit or a point on a garden trail.

  2. Step 02

    Script drafted

    Claude shapes the notes into a spoken script of an agreed length, keeping every fact from the notes and adding nothing new.

  3. Step 03

    Script approved

    A person checks the script for accuracy, names and pronunciation, and edits it before it is narrated.

  4. Step 04

    Audio narrated

    ElevenLabs narrates the approved script in your chosen voice, and a person listens to the recording before it goes live.

  5. Step 05

    Stop published

    The audio and a written transcript are published on a WordPress page, reached by a QR code at the stop.

The result

Every visitor hears the story, even with nobody on hand to tell it.

Nothing is published until a person has approved the script and listened to the audio. Voices come from a licensed library, or are cloned only with the speaker’s recorded consent.

Before we connect it.

Where this is useful

Useful for small museums, heritage sites, gardens, galleries and visitor attractions that would like an audio tour but can’t justify studio recording every time something changes. A sensible first version covers a short trail of a few stops in one language, so you can hear how visitors respond before adding more.

What we need to connect

  1. 01Notes for each stop, written or checked by someone who knows the collection, with names and their pronunciation marked.
  2. 02An ElevenLabs plan that allows commercial use, Claude API access and a WordPress site you can add pages and audio to.
  3. 03The voice you want, the length of each stop and whether you need other languages or a simpler version for families.

What needs careful handling

Synthetic voices can mispronounce place names, family names and technical terms, so pronunciation is checked during approval and corrected with spelling hints, or a pronunciation dictionary where the tool supports one. Phone signal can be weak in old buildings and gardens, so audio files are kept small and on-site Wi-Fi is worth considering where signal is poor. Every stop has a written transcript, which helps deaf and hard of hearing visitors and anyone without headphones.

Questions people ask.

Will visitors know the voice is AI-generated?

We think they should, and a short note on the tour page says so. Being open about it is fair to visitors and avoids any suggestion that a real person recorded it.

Could we use the voice of someone from our team?

Yes, if they agree in writing and ElevenLabs’ own consent and verification requirements are met. Their agreement should cover how long the voice is used and what happens if they leave or change their mind.

Stay confirmed → Host replies
Automation 134

Guest welcome app for B&Bs and holiday lets

Appointment found → Display reset
Automation 03

Calendar-triggered TV welcomes

All automation examples